We’re still waiting for free agency and teams to decide who they will use the franchise tag on, but the NFL season is 12 months long these days. With that in mind, I wanted to look at how well previous seasons’ results predict the following season.
I first did a simple correlation between the current season and the previous season. Then the current season and the average of the two previous seasons, and so on. Correlation works on a -1 to +1 scale. A correlation of zero means that there is absolutely no correlation between the two sets of data. A correlation of +1 means that the two data sets are 100% correlated.
No matter which data set you compared, all of them showed a positive correlation. However nothing was even close to having a correlation near one, which is expected given the roster turnover, injury rates, and dumb luck involved in a 16-game football season.
That being said, I wanted to take it a step further and run a regression using previous wins as well. In this instance I used the three previous seasons as variables to predict the outcome of the current season. Using the data from 2012-2014, I listed the projected 2015 win total alongside the actual win total for this past season.
|Team||2015 Proj||2015 Actual|
While some of these totals look silly after the fact, if you compare the projected totals to the actual season win totals from various sportsbooks, you will see that most of them are not far off. With that being said, the R-squared value for the 320 observations is 17.7%. This means that there is still a lot of uncertainty after this regression.
For example, based on the 2012-2014 seasons, the 2015 49ers projected win total according to the regression is 9.2 wins. CG Technology opened the San Francisco win total at 8.5 and months later, it was down to 6.5. The regression was only looking at previous seasons and didn’t take into account the facts that Jim Harbaugh had left the team for Ann Arbor, or that seemingly half of their defense was lost to either retirement or free agency.
So there is the badly needed caveat, but I want to show the 2016 projections based on the regression as well.
The teams with exceptional quarterbacks I expect to be higher than the projections above. The Patriots, Packers, and Steelers are all public favorites and will likely open a win or two higher than projected. On the other end, teams like Tennessee and Cleveland will open lower as not many bettors will be eager to bet the over on these inept franchises. There are also teams greatly affected by quarterback decisions. We don’t know who will be the starter for Denver next year and with Romo back under center in Dallas, you can expect the Cowboys total to be higher than eight.
Soon we’ll have actual win totals to compare to and as free agency unfolds, we’ll see how the market reacts to them. After all, the NFL season never ends.